Rules and Norms: Requirements for Rule Interchange Languages in the Legal Domain

Notes

mostly focused on representing legal documents rather than modeling directly legal rules.

NOTER_PAGE: (2 . 0.2139406487232574)

comprehensive list of requirements for devising rule interchange languages.

NOTER_PAGE: (2 . 0.7039337474120083)

determinative rules, which define concepts or constitute activities that cannot exist without such rules.

NOTER_PAGE: (3 . 0.3712905452035887)

technical rules, which state that something has to be done in order for something else to be attained

NOTER_PAGE: (3 . 0.4085576259489303)

prescriptions, which regulate actions by making them obligatory, permitted, or pro- hibited

NOTER_PAGE: (3 . 0.4451345755693582)

Important requirements for legal rule languages

NOTER_PAGE: (3 . 0.7395902181097158)

Reification [15]. Rules are objects with properties

NOTER_PAGE: (4 . 0.1394064872325742)

An interesting question is whether rule interchange languages for the legal domain should be expressive enough to fully model all the features listed above, or whether some of these requirements can be meet at the reasoning level

NOTER_PAGE: (6 . 0.18771566597653555)

The Rule Markup Language (RuleML)

NOTER_PAGE: (6 . 0.37405106970324364)

expected that RuleML will be the declarative method to describe rules on the Web and distributed systems

NOTER_PAGE: (6 . 0.6031746031746031)

currently there is no dialect specifically intended for the representation of legal rules

NOTER_PAGE: (7 . 0.12077294685990339)

The contribution of [21] by Grosof was the proposal of adopting courteous logic programming (a variant of defeasible logic) as execution model for RuleML rule-base corresponding to the clauses of a contract.

NOTER_PAGE: (7 . 0.1932367149758454)

The limitation of [21] is that it does not consider normative effects (i.e., it is not possible to differentiate between obligations and permissions). This limitation has been addresses by Governatori [16], where defeasible logic is extended with the standard de- ontic operators for obligations, permissions and prohibitions as well as a new special deontic operator to model violations and penalties for the violations.

NOTER_PAGE: (7 . 0.34038334434897555)

Semantics of Business Vocabulary and Business Rules (SBVR)

NOTER_PAGE: (7 . 0.6159947124917382)

focus of SBVR is on modelling not providing a framework for executing the rules.

NOTER_PAGE: (7 . 0.7937871777924652)

normal deontic logics –that is logics that admit necessitation (i.e., from  φ derive  Oφ , where O is the deontic modality for obligation)– are not suitable to model norms.

NOTER_PAGE: (8 . 0.21282220753469927)

being based on classical first order logic it is not suitable to handle conflicts.

NOTER_PAGE: (8 . 0.5194976867151355)

The Semantic Web Rule Language (SWRL)

NOTER_PAGE: (8 . 0.619960343688037)

While both OWL-DL and Datalog, separately, are decidable subsets of first-order logic, the union OWL-DL and Datalog, as in SWRL, is undecidable.

NOTER_PAGE: (8 . 0.7091870456047588)

Since SWRL rules are Horn clauses, it is not possible to model legal rules in an isomorphic way.

NOTER_PAGE: (9 . 0.17911434236615995)

lack of negation in Horn clause logic is a problem

NOTER_PAGE: (9 . 0.21414408460013218)

SWRL inherits the mono- tonic semantics of classical first-order logic

NOTER_PAGE: (9 . 0.34170522141440846)

The Rule Interchange Format (RIF)

NOTER_PAGE: (9 . 0.6318572372769332)

RIF Core and RIF-BLD both appear to suffer from the same problems as SWRL, and for the same reasons.

NOTER_PAGE: (10 . 0.2643754130865829)

An interesting question is whether the RIF Framework for Logic Dialects (RIF-FLD) could be used to con- figure a RIF dialect which is more suitable for modeling legal norms. Although this question requires further research, our initial impression is that the space of configura- tions possible is limited

NOTER_PAGE: (10 . 0.3562458691341705)

The Legal Knowledge Interchange Format (LKIF)

NOTER_PAGE: (10 . 0.5723727693324521)

The goal of the ESTRELLA project, with respect to LKIF, was to develop an interchange format for formal models of legal norms which is sufficient for modeling legal knowledge in a broad range of application scenarios, builds on existing standards, especially in the context of the Semantic Web, and informed by the state-of-the-art of the field of Artificial Intelligence and Law.

NOTER_PAGE: (10 . 0.6318572372769332)

LKIF is an XML Schema for representing theories and arguments constructed from theories. A theory in LKIF consists of a set of axioms and defeasible inference rules.

NOTER_PAGE: (10 . 0.7217448777263714)

Axioms are named formulas of full first-order logic.

NOTER_PAGE: (10 . 0.8519497686715135)

The schema for atomic formulas has been designed to allow theories to be displayed and printed in plain, natural language, using Cascaded Style Sheets (CSS).

NOTER_PAGE: (11 . 0.21414408460013218)

While LKIF includes a sublanguage for first-order logic, LKIF has been designed to allow formulas of first-order logic to be represented in human readable form in natural language

NOTER_PAGE: (11 . 0.8433575677461995)

ISO Common Logic standard does not look like it will be widely adopted within the World Wide Web community

NOTER_PAGE: (12 . 0.08460013218770654)

A reference inference engine for LKIF, called Carneades

NOTER_PAGE: (12 . 0.3218770654329147)

Since LKIF is a very expressive language, the computational complexity of various reasoning tasks can be high

NOTER_PAGE: (12 . 0.41308658294778583)

legal rea- soning and argumentation necessarily has a procedural component.

NOTER_PAGE: (12 . 0.5928618638466623)

legal rules are often defeasible and cannot be cor- rectly represented through material implications in first-order logic. Hence, LKIF and RuleML look suitable and more flexible in this regard.

NOTER_PAGE: (13 . 0.25049570389953735)

crucial for modeling legal rules is the correct representation of the many different types of normative effects, and the need to capture, for example, the deontic concepts. Here, SBVR and RuleML [16], though with some limitations, show how to do that in a rather satisfactory way.

NOTER_PAGE: (13 . 0.30601454064771977)